The Struggle to Find Expected Value in AI Investments – Techstrong AI Podcast EP59

Hello and welcome to the Techstrong AI podcast. I'm Amanda Ani, and with me today I have James Rold. He's the head of Turing Intelligence at Turing.

How are you doing today? I am fantastic. Thank you for having me.

Happy to have you on the show. Um, so can you share a little bit about touring and what services do you provide? Absolutely.

So for those of you who aren't familiar, touring is a a i infrastructure company that does two major things. One, we work with the large frontier models, you know, opening eyes, philanthropic, Geminis the world to help advance all their models and help push AI forward, say, advancing it. Then two, and what my business is, how do you then help deploy it?

So we now work with a forced foreign company and say, Hey, all these advancements in LLMs or agent flows, or all the different things that these amazing companies are creating, what do we now do to deploy that to make your company more effective? Maybe it's supply chain, maybe it's demand forecasting, maybe it's a agent, this or that. And so basically all this amazing technology over here, how do we bring it to you to make your company more successful?

Wonderful. All right, well, our topic for today is, um, implementing AI tools and finding that value and return on investment. Um, so from your experience, where are companies falling short where they are not seeing this return on investment?

Where's the critical errors that are being made? I think that the biggest one right now is I bought chat GBT for everyone, or I bought philanthropic for everyone. You know, I, I bought enterprise versions of one of these products and I've now have an AI strategy, and they're like, oh, cool.

I, I gave it to my people. Now I'm an AI company. And that's not really how it works.

And so I think kind of, I think there's, there's two major gaps. One is a, is the leadership bought into then, you know, drive all the optim. And two, some of the, some of the power of AI is definitely gonna be bottom up where I, as individual, you know, you are preparing to meet with me, you can do some research faster, I can write my blog post faster, I can do things faster.

That's definitely true. But that's the kind of what I'll call the bought up version, the tops down version, the workflow version is there are all these flows we have in our companies, invoice processing, all these different flows. It could be ENC, coding, whatever it might be, where there's just a lot of manual work.

And so I think it's what companies need to figure out is like, which flows are better off being done by AI or through automation and which, and for those, rather than just giving people better tools, you should have the tool do most of the work. And the people, you know, the quote, human in the loop can come in and make sure that the LLM or the agent is doing it well. So I'm not saying yet, yet everything is kind of completely handed off to ai.

That's not, we're not there yet in most areas, but the, I gave my people the tool, they can now go, you know, fish, I, I gave 'em a rod, they can go, fish doesn't really work. There's a lot of change management. There's also a lot of like, how do I make sure the flows I'm giving my team are AI powered or AI forward and then they can much more effective.

And so I think kind if the, yeah, you have chat VT your ai, right? That's not how, I mean, a by the way, chat's fantastic, but it's like, you know, necessary but not sufficient is the way I think about it. Mm-hmm.

So what tips do you have for business leaders to ensure good communication and proper training of how to use, uh, these AI tools and how to implement them properly? I think it starts with what is the business objective. And so, you know, if I think about, you know, my, my, I'll set up my job because then we can think about, you know, how that would work.

You know, for touring, you know, touring's job is to build a thriving business, again, helping, whether it's the ai, AI models themselves or the companies deploying ai. So we have business problems and then we're trying to figure out internally, how do we now code more effectively? Okay, great.

That's where a, a wind surf or a cursor or with the different solutions we use there might be helpful. Or we're trying to figure out how to do ticketing internally for, you know, HR issues. Great.

We've built a bunch of agentic flows there. And so again, I think the, the biggest thing is, yeah, I guess to your earlier question, like e either I've given you ai, now you're done, or start with the ai, AI is cool, you should use it. Versus like, what's the business problem I'm trying to solve?

What is the workflow I'm already doing? And then how do I do that, you know, faster, cheaper, better or, or more of it. By the way, what's AI now is where you couldn't necessarily scan every email or scan ev have every single person get the same customer experience.

Now with ai, you can all get a personalized experience. It used to be, you know, you had to be a very high value customer to get to talk to some concierge and get kind of personalized experience with ai. Everyone can.

And so this whole idea of the connection between what is happening, like obviously, which we, we had touring, see on the frontier models front back to what is the business problem you have and whether that could be at the CEO level or CXO level where you understand what the business problem is, but then like, how do these technologies over here make this business problem over here work better? So I think, so I think kind of, yeah, either giving it to bottoms up or saying please just use ai, but I don't know why I'm using ai. So again, the, the biggest thing is, you know, I I, I haven't changed my flows.

I don't do different things necessarily because of ai. I do the same things that I would've done before way, way better, or of course I use AI to outsource some of these things. But the business problem we all have, I'm trying to build a business, you're trying to build a business.

We're all trying to build business. We're all trying to drive organizational change, whatever it might be, start there and then move to how does AI accelerate that? What does the AI forward version of that look like?

Not here's the tool. You know, going inside out from technology out to use cases doesn't work as well as what are the use cases back to technology. So from your experience, uh, or Turing's experience working with some companies, can you give some actual, um, use cases or some examples of successful, um, AI adoption?

Yeah, for sure. So I think there's a bunch of use cases around auditing, whether that auditing could be for insurance companies, whether that's an, or medical auditing, where there's a lot of this information coming through, whether it's, you know, a claims or it's a, here's a, uh, you know, could be car insurance, it could be medical insurance, whatever. And so we have all this volume information where most of the information's coming through is gonna be in text format, somewhat structured wet, always that structured.

And so what's happening right now is I, as a, uh, a reviewer, as a human reviewer need to go through and check all these things. Or maybe it's similarly back to your question around like invoice processing. We all know invoices look like we all purchased orders, we will, but they all look a little different.

So humans have spent all this time looking through and like making sure this is in the right place and connecting this piece of information here to this piece of information here. What AI can do, and we're seeing with our clients is it takes these processes that are extremely manual and not particularly skilled. Not, not like low skill, but not high skill either and says, I can get you a 95% is good version in, you know, seconds that might have taken a human minutes or even hours.

And then again, back to that point earlier, human loops very important over time, AI may reach the point where it's quote better than humans, where actually, you know what? The AI solution is consistently better than humans. I don't need to have a human in the loop.

We're not there yet, the most part, especially when the co you know, when things get complicated. And so right now it's more like AI says, I think this is the answer, and the human says yes, that that's correct, and it takes him or her five seconds when before it might've taken him or her, you know, five minutes or, you know, potentially even into hours depending on how complex the, the, the document might be. And so I think the, the companies that I can identify here are the use cases of my company that are repeatable, you know, probably could be low scale, could be medium scale over time, it could be high scale too.

And then figuring out for these use cases, how much of this requires real human nuance? How much of this more like if I have a really powerful system, I can get to either a good enough or a even or a better versioning. And that's why I think the, the curve is so important because today, again, human loop's gonna be the main thing in a year or two as these frontier models we work with closely get better and better and better.

The, the bar where, you know, humans need to get involved will keep moving up and up and up. Which to a point where for some of these flows, the AI answer be better than the human answer would be. So how, what advice do you have for how business leaders can actually track their AI investments and and determine if they've received the value they were expecting?

I think kind of like you would to been, imagine we're having this conversation pre ai and you'd say, we're thinking of buying this new tool, this new software, or we're trying to work, we we're hiring a consulting firm, do some work. They're gonna figure out that we believe this problem is, is yay big. Maybe we thi we think this, you know, revenue and growth over here or cost savings over here.

We think it could be x. It's the same math of people making ROI decisions for a long time. Now, the only difference is the, the tool you're buying is an AI tool or the service you're buying, kinda like touring is an AI enabled service, but it's, it's not, it's not a new muscle of, you know, you and I make decisions all the time is, is this worthy of my spend?

Whether it's, you know, as a human, as a, as a consumer, should I buy this car or this phone or as a business person, should I buy this piece of software or even hire this person for this role? We spend our whole, you know, days at work making ro i decisions, whether it's ROIs for spend or how we spend our time. AI doesn't change that.

Again, it comes back to the business problems don't change. It's which business problems can AI help you solve? And therefore if I believe it can help me solve the problem, how big was the problem?

And therefore how much money is it worth my investing in solving that problem? Do you think that companies still struggle from employees using their own AI tools outside of company provided tools? And is there an issue there as far as security?

Yes, that definitely happens. I do think, you know, back to the earlier point, open AI and philanthropic and others have, you know, built pretty large businesses pretty quickly on both also the consumer side, like you mentioned on the enterprise side, I think, you know, increasingly more and more companies are realizing, I want all my employees to have access to these incredible tools, therefore I'm going to purchase chatt for the most part for my employees. And obviously they're gonna encourage their employees to build custom GPTs.

There is still exactly your point, some employees who are like, oh, I, I, I prefer Gemini and the company hasn't bought Gemini 'cause we're an Azure shop or we're, you know, a Microsoft office shop and therefore I'm gonna keep using Gemini. And that's potentially a problem. And I think, uh, I think over time though more, you know, I can't, you know, how everyone today has a, has a seed of Microsoft office or a seed of, you know, G Suite or Google, you know, Google's products, et cetera.

I just don't think there's a world in a couple years where every single employee doesn't have access to these tools 'cause they're so powerful. And therefore I think the, the risk doesn't disappear. 'cause again, it could be, again, I'm a Microsoft office shop, but I prefer Claude, but it at least minimizes.

And so, uh, that's something that we're seeing, you know, much more adoption of the enterprise products office. That's where they come in, that they first buy the, the core products, now they come to us. So they come to product companies like us and say, great, we've now invested in this kind of foundational layer of, um, of the user level.

How do we now automate the workflows on top of that to make that kinda the base layer more powerful? Alright, so let's look toward the future. So a couple years from now, as rapidly as AI is advancing, uh, what do you envision as far as AI in the enterprise?

I think it'll become just much more native. And so I think you think about mobile. I remember I was at LinkedIn for a long time before I was a touring.

I remember in the early days of mobile, 20 11, 20 12, 20 13, we had a mobile team. So, you know, everyone's like, oh, what's your mobile strategy? No one in 2025 has a mobile strategy, if that makes sense.

They just have a company strategy. And so I think to your question right now, in 2025, in the next few years, companies definitely need to have an AI strategy and people have, you know, AI transformation leaders or head of ai, you know, strategy, et cetera, that's gonna go away. And the best business leaders will just be the best business leaders.

And that that, you know, obviously you won't be a great business leader in 2026 or 2027 or 20, 20 30 if you don't speak fluent AI in the same way. You couldn't have been a great product leader at a Google or a Facebook or LinkedIn or whatever if you didn't speak fluent mobile. And so I think it's one of those things where it just becomes, we, we stop thinking about it as AI and we start thinking about it.

It's like, oh, now I have capabilities that can do these incredible things. And that's where, you know, companies obviously come to us because we're, you know, in, in 10 years time, maybe AI is like a core thing Today. It's so new and it's so innovative that our ability to work with the open AI philanthropics of the world to see where the future is going means we can now kind of look around corners and figure out, okay, great, we know where the future's going on the technology side for you.

That means there are opportunities to drive revenues over here, increase customer satisfaction over here, reduce costs over there. And so it's, uh, it's a, it's, it is obviously, it's very, very exciting, uh, to be, to be in this space and beginning the space will quote go away as it just becomes part of like what we're all used to. All right.

Well if there was one key takeaway you could leave our audience with today, what would that be? I think it's just start with the business problem. Starting with the AI and trying to figure out how to use it to me is never as helpful as start with whatever the problem is.

Again, whether it's a CXO trying to solve is a most important problem, or you as an individual making your day more productive. Start with what, what are your user flows? What are you trying to do?

What's the job you're trying to accomplish? And then given that, how do I use, whether it's a off the shelf product like Aachi or Alaw or increasingly as companies rely on companies like Turing to build these more automated workflows, how can those workflows be powered by AI to make it, you know, faster, cheaper, better, et cetera. Alright, well thank you so much for coming on the show and sharing your insights with us.

Of course. And thank you to our audience. Stay tuned.

There's more.